TOP TEAM DEMOGRAPHICS, INNOVATION AND BUSINESS PERFORMANCE ... · empirical evidence exists...

PRELIMINARY DRAFT: NOT FOR CIRCULATION OR QUOTATION 1

TOP TEAM DEMOGRAPHICS, INNOVATION AND BUSINESS

PERFORMANCE: FINDINGS FROM ENGLISH FIRMS AND CITIES

Max Nathan

LSE, SERC AND NIESR

Spatial Economics Research Centre, London School of Economics and Political Science, Houghton

St, London, WC2A 2AE. [email protected]. personal.lse.ac.uk/nathanm

January 2013

Abstract

High levels of net migration to the UK have contributed to growing cultural diversity, and researchers

are turning their attention to the long-term effects of diversity on productivity. Yet little is known

about these issues. This paper asks: what are the links between the composition of firms' top teams

and business performance? What role do ethnic diversity and co-ethnic networks play? And do cities

amplify or dampen these channels? I explore using a rich dataset of over 6,000 English firms.

Owners, partners and directors set firms' strategic direction. Top team demography might generate

production externalities through diversity (a wider range of ideas/ experiences, helping problem

solving) and/or 'sameness' (via specialist knowledge or better access to international markets). These

channels may be balanced by internal downsides (lower trust) and external barriers (discrimination),

so that overall effects on business performance are unclear. In addition, urban locations (particularly

big cities) may amplify any demographics-performance effects. I create a repeat cross-section of firms

from the RDA National Business Survey. I construct measures of diversity and sameness across

ethnicity and gender 'bases', alongside information on revenues, product and process innovation. I

then regress these measures of business performance on top team demographics, plus firm level

controls, area, year and detailed industry fixed effects. My results suggest a non-linear link between

diversity and business performance, which is net positive for process innovation and net negative for

turnover. Further tests on diverse and minority/female-headed firms find positive links for diverse top

teams, negative for minority and female-only top teams. This implies that while diversity has internal

and external benefits, penalties from being 'too diverse' probably result from external constraints.

Further tests for intervening effects of capital cities, metropolitan hierarchies and urban form find

some evidence of amplifying and dampening effects – which are generally stronger in London and

larger cities.

JEL classification: J61, L21, M13, O11, O31, R23

Keywords: cities, innovation, entrepreneurship, cultural diversity, migration, gender

Acknowledgements

The research is supported by LLAKES, and makes use of microdata from the RDA National Business

Survey. The findings are those of the author, not those of the funders or data providers. Many thanks

to Patrick Bowes (ex-Yorkshire Forward), Richard Woolhouse and Jonathan Wood at the CBI, and

Adam Swash and Sachin Dodhia from Experian for use of the data. Thanks also to Cinzia Rienzo for

helpful comments on earlier drafts. The opinions in this paper are those of the author, not those of the

funders or data providers. All errors and omissions remain my own.


1. Introduction

This paper asks: what are the links between the demographic composition of senior staff in

firms, and those firms' levels of innovation and revenues? What roles might ethnic and

gender diversity have? Are minority- and female-headed firms at an advantage? I use a rich

dataset of nearly 10,000 English firms to explore these issues.

These questions are important for both researchers and policymakers. The UK, like

many other Western countries, has become substantially more ethnically and culturally

diverse in recent decades, with net migration a main driver. The latest Census data make this

very clear: between 2001 and 2011, the foreign-born population of England and Wales rose

from 4.6 to 7.5m (from 9-13% of the population). At the same time, the share of 'white' and

'white British' ethnic groups decreased, drops of 91.3-86% and 87.5-80% respectively.

Notably, the biggest-growing ethnic group was 'other white', with Polish-born the fastest-

growing migrant group (Office of National Statistics, 2012b, Office of National Statistics,

2012a). These demographic changes have been most striking in urban areas: notably, London

is now a ‘majority minority’ city for the first time in its history.

Given these long term shifts, attention is increasingly turning to the dynamic effects

of immigrant communities on host country economies, particularly through firm-level

channels that shape productivity, and through the diversity that migration brings (see Kerr

and Kerr (2011) for a recent review). Gender equality and diversity are also major policy

agendas in developed economies. In the UK, particular public attention is paid to the

presence and impact of women in senior positions; and to encouraging female

entrepreneurship (Davies Review, 2011).

Owners, partners and directors of firms – the ‘top team’ - help set the strategic

direction of the businesses they run, and play an important role in their success or failure

(Certo et al., 2006). In theory, there are two broad ways in which 'top team' demographic

composition might affect business performance. One argument highlights externalities from

diversity: specifically, a range of skills, knowledge, backgrounds and experiences may help

teams to generate new ideas and to problem-solve (Page, 2007). Both gender and ethnic

diversity could produce this advantageous mix. The other perspective emphasises gains from


sameness - for example, externalities from social networks or deeper specialist knowledge,

both of which may aid knowledge diffusion and market reach (Docquier and Rapoport,

2012).

Note that in theory both these channels have ambiguous effects - diverse teams may

exhibit lower trust, social networks may be constrained, and minority or female-headed firms

may experience discrimination. Thus the diversity-performance relationship may be non-

linear, with an optimal level of mix after which disadvantages outweigh advantages (Ashraf

and Galor, 2011). Note also that diversity and sameness channels are not mutually exclusive;

both could run in parallel, and that different ‘diversities may’ have different effects. What

empirical evidence exists suggests small net positive effects for ethnicity and gender on

various measures of business performance, but there remain large knowledge gaps and

problems identifying causal effects (Nathan, 2012, Adams et al., 2010, Certo et al., 2006).

In addition, cities or urban locations may amplify or dampen these processes – the

former through agglomeration or composition effects, the latter through higher levels of

competition, segregation or discrimination. These intervening factors are likely to be

particularly salient for ethnic-diverse and minority ethnic headed businesses, but little

quantitative work has been done in exploring their real effects (Nathan and Lee,

Forthcoming).

I use rich microdata from the English Regional Development Agencies' National

Business Survey to shed light on these issues. I pool data for 2008 and 2009 to create a

sample of over 6,000 firms in England. I regress measures of top team diversity and

sameness, by ethnic and gender, on measures of firms' product and process innovation

activity, and on business revenue/turnover.

My results suggest a non-linear link between diversity and business performance,

implying both positive and negative affordances. Echoing other studies, this translates to a

small net positive term of ethnic diversity on levels of process innovation. More surprisingly,

I find negative net relationships between ethnic and gender diversity and business turnover,

implying that internal or external constraints ultimately outweigh benefits. I run further tests

distinguishing diverse and minority/female-headed firms, here finding generally positive

diversity-performance links but zero or negative links for minority and female-headed


businesses. This suggests that while business diversity has (internal and external) benefits,

being ‘too diverse’ is probably an issue of external constraints to the firm (such as

discrimination) rather than internal problems. I also test for any amplifying effects of

London, other large UK metros, and urban form generally. Here my results some evidence of

amplifying and dampening effects, which are generally stronger and more visible through a

London or ‘city’ lens than a broader urban / rural one.

The paper makes a number of contributions. First, while gender diversity and its role

in ‘top teams’ has been previously explored in the empirical literature, the role of ethnic

diversity and co-ethnic communities on business performance has been under-examined,

especially in the ‘top team’ business context. Second, I look at different aspects of diversity

together, and to explore their links to multiple business outcomes – not only product and

process innovation, but resulting levels of business revenue. Third, I link the management

literature (on top teams) to research on migration and diversity issues, and research by

economists and geographers on the dynamic impacts of migration and migrant / minority

communities, especially in urban environments. In so doing, the paper adds to the European

literature on the ‘economics of diversity’ (Nathan, 2012). As far as I am aware, it is the first

quantitative study of its kind in the UK, and provides a useful extension to recent research on

diversity and innovation in London (Nathan and Lee, Forthcoming).

The paper is organised as follows. Section 2 builds a simple conceptual framework

and reviews relevant empirics. Sections 3 and 4 describe the data and identification strategy

respectively. Section 5 sets out the model and gives some brief descriptive analysis. Sections

6 and 7 describe the main results, with a tranche of robustness checks in Section 8. Section 9

discusses and sets out ideas for further research.

2. Framework and evidence

2.1 Definitions

‘Diversity’ is hard to define in a form suitable for quantitative analysis. In this context,

diversity refers to the mix of identity groups in a firm, or more precisely in the top team of


owners/partners. Identity is a multifaceted concept, with subjective elements, and categories

that alter over time (Aspinall, 2009). Gender and ethnicity are two important aspects of

identity that I will use here. Gender diversity is defined in terms of female presence.

Following the literature, I will treat ethnicity as given, not endogenous, and will largely

abstract away self-ascribed elements (Ottaviano et al., 2007, Green 2011).1 I also follow UK

Office of National Statistics ethnic group definitions, which tend to focus on ‘visible

minorities’ and operate at a fairly high level of generality. Here, ‘minority ethnic’ refers to

Black and Minority Ethnic (BME) groups.

I am interested in two main measures of business performance: innovation and

turnover. I borrow the UK Government’s definition of innovation as ‘the successful

exploitation of new ideas’ (Department of Innovation Universities and Skills, 2008).

Innovation thus involves both ‘upstream’ generation of new products and processes, and their

‘downstream’ commercialisation (Fagerberg, 2005).My data allows me to observe whether or

not this ‘upstream’ product and process innovation has taken place. Turnover is defined as

revenue: specifically, ‘turnover’ is the level of revenue that a firm receives from its normal

business activities in a given time period.

The notion of a ‘top team’ is taken from the management literature, specifically the

‘upper echelons’ research pioneered by Hambrick and Mason (1984). The definition is

deliberately broad to cover the ‘dominant coalition’ in a firm, comprising both senior

directors (who may be employees) and owners and partners (Carpenter et al., 2004).

2.2 Framework

There are two main perspectives on how top team demographics may affect business

performance. The first view emphasises the importance of diversity. Diverse firms and teams

may benefit from a wider range of ideas, perspectives and backgrounds, which ought to

improve problem-solving and ideas generation – and thus raise measures of innovation to the

firm (Page, 2007, Berliant and Fujita, 2007). Diversity may also help firms to better handle

complex external business environments (Williams and O'Reilly, 1998). In both cases,

1 If identity is entirely self-ascribed, it becomes very hard to link behaviour to measures (Casey and

Dustmann 2009). However, in practice it unlikely that (for example) commercial success might lead

business owners of South Asian origin to identify as ‘White British’.


demographic structure should feed through into higher revenues. These effects may be

particularly important in ‘knowledge-intensive’ settings (Fujita and Weber, 2003).

Conversely, diverse firms/teams may face internal challenges – specifically, trust and

bonding social capital may be lower than for homogenous groups (Alesina and Ferrara,

2005). And externally, such firms may face discrimination from customers or suppliers (for

example, finance providers). Both of these forces will have a negative influence on

innovation and revenues.

The second view focuses on dimensions of ‘sameness’. In part, negative affordances

of diversity are simply positive affordances of similarity. However, theory also suggests

further externalities that benefit firms. For example, co-ethnic networks may reduce

transactions costs and aid knowledge diffusion (Agrawal et al., 2008, Docquier and Rapoport,

2012). Identity group membership may aid market access either geographically, through

diasporic communities, or in terms of product space – for example, female-headed firms

probably have better market knowledge of products and services aimed at women and

families (Javorcik et al., 2011, Foley and Kerr, 2011). These channels should aid innovation

and revenue growth respectively, and may be particularly important under globalisation

(Saxenian, 2006, Yeung, 2009). However, sameness may also have downsides. Within the

firm, a lack of diversity may shut off sources of innovation stemming from unfamiliar

perspectives or knowledge (Boschma, 2005); externally, minority-ethnic or female-headed

businesses may experience discrimination, limiting the ability to commercialise innovations

and constraining revenues (Zenou, 2011, Patacchini and Zenou, 2012).

This brief discussion raises a number of points. First, both diversity and ‘sameness’ in

firms may influence business performance. Equally, both have pros and cons, so predicted

effects are ambiguous. Second, diversity and sameness operate through both distinct and

overlapping channels, so could be complements or substitutes. Third, the shape of the

relationship to performance is unclear: for instance, given the pros and cons of diversity, the

true diversity-performance relationship in a firm/team may be U-shaped rather than linear, so

that an ‘optimal’ level of ethnic/gender diversity exists (Ashraf and Galor, 2011). Fourth, it is

important to look at different identity bases in isolation. Different aspects of diversity /

sameness, such as gender and ethnicity, might then be linked to different outcomes.


It is also important to consider how these channels may operate in different parts of

the firm. The demographics of senior management and the wider workforce may have

different effects on measures of business performance. In theory, ‘top team’ composition is

likely to be highly important: senior managers set the overall direction of the business, take

strategic decisions and tend to have the most experience and human capital. Beginning with a

seminal paper by Hambrick and Mason (1984), a number of studies in the management

literature have developed models of firms’ ‘upper echelons’ or ‘top management team’

(TMT), where the size, structure and composition of senior management have important

direct and indirect effects on business performance (see Certo et al (2006) and Carpenter et al

(2004) for recent reviews, and Adams et al (2010) for a related and highly relevant discussion

on corporate boards). TMT models highlight the critical role of team ‘demographics’ as

observable proxies for behaviour: demographics are defined broadly to encompass age,

gender, race/ethnicity, human capital, function, background and degrees of internal / external

/ international corporate experience. TMT models also highlight important ‘intervening’

factors both at firm level and in the wider business / social environment.

Spatial context provides a further dimension to explore. Specifically, cities or urban

locations may amplify or dampen these processes – the former through agglomeration or

composition effects, the latter through higher levels of competition, segregation or

discrimination (Jacobs, 1969, Gordon et al., 2007, Berliant and Fujita, 2009, Goldin et al.,

2011, Zenou, 2011). These intervening factors are likely to be particularly salient for ethnic-

diverse and minority ethnic headed businesses, but little quantitative work has been done in

exploring their real effects, especially in the UK (Nathan and Lee, Forthcoming).

Despite the richness of these frameworks, identifying causal effects of diversity /

sameness on business performance presents a number of major challenges, which I briefly

preview here. A first issue is to try and isolate team/group-level effects from individual

characteristics, other firm-level characteristics and wider contextual factors deriving from

industry, time trends, local area conditions or policy shocks. A second is the chance of

simultaneity or causation at area level; successful firms may select into the largest markets,

which ceteris paribus will tend to have larger and more diverse populations. A third issue,

which is very hard to disentangle, is both-ways causation within the firm. Suppose there is a

‘diversity bonus’ of some kind which positively influences company performance; firms

observe this and change their hiring patterns to suit. In practice, I am able to deal with the


first and second issues through careful controls and fixed effects; future versions of the paper

will use instruments to deal with the third. There is further discussion in Section 5 below.

2.3 Evidence base

The existing literature on these issues falls into two broad categories. Economists

studying migration issues are increasingly trying to analyse links between migration, migrant

communities and productivity – at individual, firm and area level. Many of these studies also

look at second and third-generation communities, and more broadly at economic impacts of

ethnic diversity.

A handful of these studies look specifically at diversity and business performance at

the firm level, focussing on innovation outcomes. Ozgen et al (2011) find some positive links

between migrant worker share, workforce diversity and innovation in knowledge-intensive

Dutch firms. In Denmark, Parotta and colleagues (2011) find significant positive effects of

cultural diversity on firms’ propensity to innovate and on productivity – but again, only in

‘white collar’ sectors employing predominantly skilled workers. Lauren et al (2004), in a

study of engineering consulting firms, find a curvilinear relationship between human capital

diversity and business performance. Maré et al (2011, 2011) find no systematic links between

workforce characteristics and innovation, but some productivity links, among businesses in

New Zealand. In the UK, Nathan and Lee (forthcoming) find positive links between top team

diversity and innovation in London firms.

A larger number of studies in look at diversity and market orientation (see Page

(2007) for a recent review). For example, in a study of 165 Swiss firms, Nielsen finds that

nationality mix in management teams is linked to higher rates of foreign market entry and

greater profitability (cited in Hart (2010)). International evidence from economic geography

also suggests that diasporas can engage in innovative activity. Saxenian (2006) and Saxenian

and Sabel(2008) provide detailed evidence on the roles of migrant diasporas in Silicon

Valley, which have strong links to production clusters in India, Taiwan and (increasingly)

China. Similarly, Kapur and McHale (2005) and Kerr (2008) detail the roles of diasporas in

the development of ICT clusters in Ireland, Israel and South East Asia. Dahlman (2010)

shows how national Governments in BRIC (Brazil, Russia, India and China) countries have

taken an increasingly active role here.


Notably few studies in this tradition attempt to look at multiple diversity bases, and

very few focus on senior personnel. An exception to the former is a cross-sectional Danish

survey by Ostergaard et al (2011), which finds no significant links between ethnicity and

propensity to innovate, but a positive link between an ‘open’ firm culture and innovative

performance, and a positive association between firms’ gender diversity and the propensity to

innovate.

In contrast, the strategic management literature has a long tradition of empirical

‘TMT’ research, and analysis on multiple aspects of workplace diversity. Carpenter et al

(2004) and Certo et al (2006) provide useful reviews of the TMT literature and conduct meta-

analyses. Both find that while there are typically modest effects of top team demographic

factors on business performance, there are substantial intervening elements both at firm level

and in the wider industry / spatial environment.

Individual studies in this tradition predominantly focus on teams’ mix of age,

education, function and background (see for example Buyl et al (2010), Naranjo-Gil et al

(2008), Pitcher and Smith (2001), Wiersema and Bantel (1992), Bantel and Jackson (1989)).

Jackson et al (2003) review the wider workplace diversity research: they note that while

gender-based analyses are relatively common, ethnicity-based studies are much rarer, and

there are very few studies which attempt to combine ‘multi-dimensional diversity’ in a way

that reflects actual processes of self-ascription, e.g. ‘Asian female scientist’.

A much smaller number of studies look at ethnicity and/or gender mix in top teams.

For instance, Dezsö and Ross (2012) conduct a panel data analysis of the S&P top 1500

firms. They find that female representation in top management improves firm performance,

but only when the firm has an ‘innovation-focused strategy’. Asiedu et al (2012) look at US

SMEs and access to finance, finding significant differences in loan approvals and interest

rates between firms owned by white males and those owned by minority or white females.

Francoeur et al (2008) suggest that firms operating in complex environments generate

positive and significant abnormal returns when they have a high proportion of female senior

managers. Dahlin et al (2005) find that national diversity in teams has a u-shaped relationship

with information sharing and use.


3. Data

My main data source is the English Regional Development Agencies’ National Business

Survey (hence NBS), which was conducted in two waves every year from 2003 through to

2009 (the Agencies were formally abolished in 2011). Each wave covered around 5,000 firms

across the nine English regions including London.2 Data has been weighted by employee

numbers and region, to reflect the national profile (Ipsos MORI, 2009). The NBS included

questions about owner/partner ethnicity and gender in the Autumn waves of the 2008 and

2009 data, and these form the basis of my sample.

The NBS has many strengths. The UK has surprisingly few rich sources of firm-level

data. The NBS is a single source that asks a detailed range of questions about business

performance and constraints, as well as top team and firm characteristics. Importantly, the

data allows me to separately identify diversity and sameness information along multiple

dimensions, alongside multiple measures of business performance. The NBS also includes

industry codes at up to four-digit level and detailed spatial identifiers for NUTS1-3 areas,

enabling me to fit detailed sectoral and area fixed effects alongside firm-level controls. As

such it is substantially more informative than other business-level datasets such as the ARD,

and more comprehensive in its issue coverage than survey-based data such as the Community

Innovation Survey or the Workplace Employer Response Survey.

However, there are also limitations to the data. It is a sample rather than a universe of

firms, and information on ethnicity is only available for a couple of years. There is no panel

structure to the data, so a repeat cross-section is the only feasible setup. In some areas of the

survey the question format also varies significantly from year to year, so that constructing

time-consistent variables loses some detail available in individual cross-sections. Finally, the

NBS only contains information on top team demographics, rather than the wider workforce,

and has no direct information on senior individuals or firms’ wider human capital. To deal

with this last issue, I draw on detailed small-area level human capital and occupational

structure information from the Annual Population Survey (APS), which contains a boosted

2 The full list of regions is the North East, North West, Yorkshire and Humber, West Midlands, East Midlands,

East of England, South East, London and the South West.


local sample which allows for reliable sub-regional estimates.3 To enable better exploration

of urban / city-level effects, I also add in Eurostat and ONS typologies of urban-rural form

and metropolitan hierarchy (see section 7 for more details).

4. Identification strategy

I use the NBS to explore links between measures of top team composition and measures of

business performance. Specifically, I link firm-level variations in ethnic and gender team

composition to variations in firms’ turnover and innovative activity, while controlling for

other firm, industry, area and time characteristics. I am particularly interested in 1) whether

an increase in senior management diversity is linked to an improvement in business

performance; 2) whether diversity and sameness are substitutes or complements, and the

relative size and direction of their effects, and 3) which dimensions of diversity (sameness)

matter, that is, the relative roles of ethnicity and gender as ‘bases’.

I construct the sample by combining the 2008 and 2009 Wave 2 cross-sections. I

restrict the analysis to firms for which there is information on innovative activity, turnover,

industry and area, giving me a basic sample of 6,227 observations. Each observation

represents a single firm coded to one of 62 two-digit industry categories, geocoded to one of

107 NUTS3 areas and observed in a single year (2008 or 2009).4

4.1 Main variables

My independent variables of interest are measures of top team diversity and

similarity, using ethnicity and gender bases. The NBS provides information on the ethnic and

gender composition of firms’ owners / partners / directors. I use this to build two types of

variables covering diversity and sameness. First, I make continuous variables measuring a)

3 The Annual Population Survey (APS) combines results from the English Labour Force Survey (LFS) and the

English, Welsh and Scottish LFS boosts, and asks 155,000 households and 360,000 people per dataset about

their own circumstances and experiences regarding a range of subjects including housing, employment and

education. The APS’ increased sample size provides substantially greater precision than the LFS when working

at sub-regional level, as the analysis in this paper requires. 4 I explore various cell configurations, covering SIC1-4 industry codes and NUTS1-3 area codes. My aim is to

get the richest area and industry fixed effects without inducing measurement error through small cell sizes. In

robustness checks I a) drop cells with NUTS2 and SIC3 frequencies under 10 b) use SIC1, NUTS2 and NUTS1

fixed effects with very little change to the main results.


the share of minority ethnic owner/partners in the firm, and the b) share of female

owners/partners. These are my basic measures of diversity. I also construct c) quadratic terms

to explore the potentially non-linear relationship between diversity and performance.

Next, I make a series of dummy variables for both ethnicity and gender,

distinguishing firms with all majority ethnic (white British) owners/partners, and all-male

owners/partners (‘homogenous firms’), a mix (‘diverse firms’) and all minority ethnic /

female owners/partners (‘minority ethnic-headed’ / ‘female-headed’ firms). This allows me to

fit measures of diversity and sameness together, testing whether the two are complements and

substitutes. More broadly, I am able to look at the degree of complementarity between

ethnicity and gender bases.

My dependent variables are innovative activity and revenue / turnover, which are also

well covered in the NBS. For innovative activity, I fit dummies taking the value 1 if the firm

has, in the past 12 months, introduced 1) a new product innovation or 2) a new process

innovation. These definitions are deliberately broad, as survey-based analyses need to capture

very different innovation conditions across manufacturing and service sector firms.5 Annual

turnover information is provided in bands – eight bands in 2008, and four bands in 2009

(<£100k, £100-999k, £1-5m, >£5m). For the full regressions I fit a time-consistent four-band

turnover variable; in robustness checks on the 2008 cross-section I use richer seven-band

information.

4.2 Identification challenges

The data structure and sample construction three main identification challenges,

which were introduced in Section 2 and discussed in detail here. The first, highlighted in the

TMT literature, is that while I want to identify group-level characteristics of top teams on

firm-level outcomes, I need to be able to isolate group-level characteristics from a) individual

group member characteristics, such as human capital and entrepreneurial ‘spirit’; b) other

firm-level factors, such as age, size and previous investments; c) wider contextual factors

such as location, time shocks, or industry trends (Certo et al., 2006, Carpenter et al., 2004).

5 An inherent limit of this approach is that it risks capturing some trivial innovations, particularly in the process

innovation category. Survey based methods may also risk a response bias towards innovating firms (Smith

2005).


Each of a) - c) presents potential intervening factors which may affect both group

demographics and business performance; for example, a technology shock might lower entry

barriers in a given industry, enabling innovation and influencing top team composition as

new firms form. Omitting these variables in regressions may lead to imprecision or worse,

spurious correlations.

I am able to deal with most cases of b) and c) using careful control variables at firm

level, as well as detailed industry, time and area fixed effects; I partially deal with a) by

fitting NUTS3-level human capital and occupational controls (see next section for more on

these). Area-level factors present a second, related problem of positive selection. Innovative

or high-turnover firms may choose to locate in the area with the greatest economic

opportunities or innovation ‘infrastructure’, and this may vary by sector and firm type

(Duranton and Puga, 2001). Not controlling for this means that coefficients of top team

composition are likely to be biased upwards. The NBS structure does not identify moving

firms, but this still leaves the possibility that historic location choices reflect persistent

differences in local opportunities. I deal with this issue through area-level fixed effects that

control for time-invariate area characteristics. I also exploit my choice of sample years: the

UK was in recession in 2008-9 and the pull of successful areas will have been dampened

during this time.

A third issue is simultaneity and/or reverse causation within the firm. If businesses

observe a positive (negative) effect of top team composition on business performance, they

may adjust team composition to maximise (minimise) any positive (negative) consequences

for the firm (Ozgen et al., 2011, Parrotta et al., 2011, Nathan and Lee, Forthcoming). This

will lead to at best, inflated coefficients of top team composition effects if not corrected, and

at worst, spurious associations. This is an issue common to many studies in this field, and

ideally, one would use a natural experiment that acted as a shifter of team composition to try

and identify causal effects (Adams et al., 2010). A final issue also identified by Adams et al

(ibid) is that, for a given company, exogenous firm-level heterogeneity may also influence

the optimal top team composition for that firm. My controls strategy should deal with much

of the observables, but the data structure does not permit firm fixed effects which would

control for firm-level unobservables. For both reasons, I interpret results as associations

rather than causal effects.


5. Estimation

I fit the data to a production-function type model, where for firm i, industry j, area a and year

t I estimate:

Yijat = a + bETEAMijat + cFTEAMijat CONTROLSijatd + Jj + Aa + Tt + e (1)

Y is variously a dummy for product or process innovation, or the firm’s turnover. Both

models relate measures of business performance to top team demographics (ETEAM,

FTEAM), a vector of firm-level controls (CONTROLS) and fixed effects.

ETEAM covers top team characteristics by ethnicity. In the main results it is the share

of minority ethnic owners/partners and its quadratic, which is my measure of diversity.

Coefficients of ETEAM reflect the joint ‘effect’ of changes in ethnic composition on Y; as

suggested above I am particularly interested in whether increases in diversity has a linear

relationship with business performance, or whether an ‘optimal’ level of diversity exists. In

extensions to the main analysis ETEAM includes dummies for minority ethnic-diverse and

minority ethnic-headed firms. This specification enables me to explore the relationship

between diversity and sameness: coefficients are ‘effects’ relative to being in a homogenous

firm, the reference category. FTEAM is organised and interpreted along the same lines but

for gender composition.

Controls are chosen to on the basis of the wider literature on business innovation and

performance. Both firm age and firm size will influence the performance of the company: for

instance, large or established firms often generate large amounts of patent activity, but small

and/or new firms may introduce disruptive innovations (Griffith et al., 2006). Young, small

firms also account for substantial shares of national output and employment growth (through

rapid scaling) (Haltiwanger et al., 2010, Biosca et al., 2011, Lee, 2012). In turn, age and size

may shape the composition of the firm’s senior team. I therefore fit controls for the number

of owners/partners, the age of the firm and the number of its employees. Company type is

likely to influence both top team demographics and corporate performance; for example,

subsidiaries and joint ventures of foreign-owned firms are more likely to benefit from

knowledge spillovers and technology transfer (Aitken and Harrison, 1999, Javorcik, 2004,


Harrison and Rodríguez-Clare, 2009). The NBS provides detailed information on company

type, so I fit dummies for UK subsidiaries, foreign subsidiaries, ultimate holding companies,

independents and LLPs (unknown status being the reference category). The NBS does not

directly include information on firms’ human capital stocks, but does ask about whether firms

have attempted to improve their skills base through internal or external training; I use this as

a proxy human capital control. I fit two controls for precision; namely dummies which take

the value 1 if the firm has a codified growth plan, and if it is operating at capacity. Both

should be positively correlated with innovation and with levels of revenue. Finally, J, A and

T represent two-digit industry, NUTS3 and year fixed effects respectively.

A number of further controls are fitted in robustness checks. First, high-performing

firms are more likely to export and work in international markets (Rodrik, 2004); supply

chain and customer market geographies may also influence the make-up of firms’ senior

management. The NBS provides information on the share of inputs sourced domestically or

abroad, and similar information for the pattern of sales, which are used to construct further

controls. However, as only a limited number of firms answer these questions they are

reserved for cross-checks.

Second, the 2008 NBS also provides information on a number of innovation related

variables, which I fit in robustness checks on the cross-section. There is an established

literature on ‘open innovation’ and collaboration, with firms that collaborate likely to access

external knowledge and produce more innovations (Von Hippel, 2005). Other studies

highlight the role of university-industry links (D’Este et al., 2011). Both of these should

influence levels of innovation, and through this firm revenue. I therefore construct dummies

for whether a firm uses specialist networks for information, and whether firms exploit

university-industry links for R&D. I also include a dummy for whether the firm expects to

invest in R&D during the year, reflecting the wide literature linking R&D and long-term

business performance (Romer, 1990).

Estimators are chosen appropriate to the data structure. Innovation models are

estimated as logistic regressions; following Angrist and Pischke (2009) I also show OLS

results to indicate marginal effects. Turnover models are estimated as fixed effects OLS

models.


5.1 Descriptives

Tables 1 and 2 provide some brief descriptive analysis. Table 1 gives summary

statistics. The first panel covers my dependent variables: under a quarter of firms have

introduced a product innovation, just under 10 percent a process innovation. Turnover is

banded in four broad categories, and suggests the average firm has a turnover of around

£100k. The second panel covers the main independent variables: the average share of

minority ethnic owners/partners is around three percent; with 2.3 percent of firms being

minority-ethnic headed. Female owner/partnership is much more common, the average firm

having nearly 26 percent female owners/partners; female-headed firms comprise just under

10% of the sample.

Table 1 about here

The third and fourth panels cover control variables. Controls in the fourth panel are used for

robustness checks; some are only available for 2008.

Table 2 about here

Table 2 shows a correlation matrix for the main dependent, independent and control

variables. Pairwise correlations are generally low, suggesting no inherent collinearity issues

in the data.6

6. Main results

Results for the main regression analysis are given in Tables 3-5 (innovation models) and 6

(turnover / revenue model). In each table column 1 fits a simple share of minority ethnic

owners / partners; column 2 adds controls; column 3 fits the share and its quadratic; column 4

adds controls to this; column 5 adds the share of female owners and its quadratic. Innovation

models are estimated in logit form, and point estimates are shown as raw coefficients. For

these models Table 5 re-runs columns 3-5 in OLS to (roughly) illustrate marginal effects.

6 Matrices for the full set of variables also suggest no collinearity. Results available on request.


6.1 Innovation results

Product innovation results are given in Table 3. The model tests the links between the

level of firms’ share of top team minority ethnic and female owners/partners, and the

likelihood that firm has introduced a product innovation in that year. The simplest

specifications (columns 1 and 2) show no linear link. Including the share of minority ethnic

owners/partners and its quadratic shows a small positive coefficient on the share, and a

slightly smaller negative coefficient on the squared term. This is suggestive of a non-linear

relationship where the joint effect is a small net positive – echoing the discussion in Section 2

– although neither is statistically significant. However, adding controls reduces coefficient

size and – surprisingly - reverses their signs. The most fully specified model (column 6) fits

shares and quadratics of both ETEAM and FTEAM. Coefficients of FTEAM are positive on

the share (0.523, significant at 5%) and negative on the quadratic (-0.618, significant at 10%).

Table 3 about here

Table 4 switches attention to process innovation. As before, fitting the share of

minority ethnic owners/partners shows no effect (columns 1 and 2), while fitting the share

and its quadratic generates a robust and marginally significant relationship, where the joint

effect is a small net positive (column 3). Interestingly, while the coefficients shrink as

controls are added back in, ETEAM remains significant at 10%. Specifically, in the most

fully specified model (column 6) the coefficient of the share of minority ethnic

owners/partners is 1.937, significant at 10%, while the point estimate on the quadratic is -

1.651. This suggests a positive link between diversity and process innovation, until a turning

point is reached around a minority ethnic ownership share of about 0.3.

Table 4 about here

The left hand Table 5 shows the marginal effects for the product innovation model,

which in for ETEAM is small and non-significant. For FTEAM, by contrast, both share and

quadratic are significant – although as they exactly outweigh each other in the OLS, the

overall marginal effect on product innovation is essentially zero (raw coefficients suggest a

small net positive marginal effect).


The right hand panel shows the marginal effects for the process innovation model. For

ETEAM, we can interpret this as showing that a 10 percentage point rise in the share of

minority ethnic owners/partners is linked to a (0.203 + (-0.178*0.1) = 0.185 probability of a

firm generating a process innovation. By contrast, coefficients of FTEAM remain

insignificant and close to zero – although note that the square of the share of female

owners/partners is (just) negative.

Table 5 about here

6.2 Turnover results

Table 6 shows results for the turnover model. Unlike the innovation models, columns

1 and 2 find a small negative association between the share of minority ethnic owners /

partners and turnover levels. Column 3 fits the share and its quadratic, and shows a large,

strong positive linear link – but a slightly stronger negative link on the quadratic. Both

coefficients are significant at 1%. Columns 4-6 add in controls. As expected, this shrinks the

point estimates but the basic shape of the result survives. In column 6, the coefficient of the

share of minority ethnic owners / partners is 0.730, significant at 5%, while its square is -

0.798, significant at 1%.

Table 6 about here

This implies that a 10 percentage point rise in the share of minority ethnic

owners/partners is linked to a (0.730 + (-0.798*0.1) = 0.068 unit fall in turnover. Note that

this result controls for the age and size of the firm, company type and some measures of firm

capacity, as well as industry, area and time fixed effects. Column 6 also adds FTEAM

coefficients. Point estimates are substantially smaller than ETEAM, but again, the joint effect

is a small net negative. In both cases this suggests a non-linear relationship between diversity

bases and company performance, with a tipping point around an ‘optimal’ diversity level.

6.3 Diversity and sameness

The main results strongly suggest a non-linear relationship between top team diversity

and business performance (and that gender and ethnic diversity play different roles).


However, this leaves open the question of whether diversity and sameness act as

complements or substitutes across the whole set of businesses. In order to explore this

further, I run further regressions distinguishing between between diverse firms (with a mixed

top team) and those headed by minority ethnic or female bosses. This allows me to look at

whether diversity and ‘sameness’ are substitutes or complements across the set of firms as a

whole – and whether different identity bases play out differently when re-cut this way.

In these models I fit dummies for ethnic / gender ‘diverse’ and minority ethnic/female

‘headed’ firms, with cofficients interpreted as relative effects of being X type of firm against

being a ‘homogenous’ firm, the reference category. Descriptive analysis in Section 5 shows

that a majority of firms are homogenous, with a minority of diverse firms and a much smaller

group of minority ethnic and female-headed businesses.

Table 7 about here

Table 7 gives results for product innovation (left-hand panel) and process innovation

(right hand panel). For product innovation, ethnic diversity has a negative coefficient and

ethnic-headed status a positive coefficient, reflecting the relationship found in the previous

results; neither is significant. Gender diversity has a positive link significant at 1%; female-

headed firm status is also positive, but much smaller and non-significant. For process

innovation, all coefficients of interest are positive but non-significant.

Table 8 about here

Table 8 gives results for the turnover model. Here, diversity measures have a strongly

positive link to turnover, with measures of minority ethnic and female-headed firms showing

negative linka. For example, the coefficient for ethnic diverse top teams is 0.165, significant

at 1%, while the beta of minority-ethnic headed firms is -0.067. For gender, respective

coefficients are 0.023 and -0.280 (1%).


7. Urban and big city location

Theory and evidence suggests that ethnicity-performance effects may influenced by large

urban environments. Cities – and urban areas more generally – may amplify these channels

(through demographic compositional effects or agglomeration economies) or dampen them

(through greater competition or demographic segregation). Ethnic-diverse firms in

cities/urban environments may therefore experience different outcomes from similar firms in

smaller, less urban locations.

In the UK context these phenomena are perhaps most likely in London (Nathan and

Lee, Forthcoming), but may also be present in other big cities and urban cores. Firm-level

demographics and urban ‘critical mass’ may therefore interact in a way not captured by my

existing control structure.

I am able to test for both city and urban effects. First, I code firms’ locations using

the Eurostat metropolitan hierarchy classification for NUTS3 areas, which sorts geographies

into four categories: ‘capital city region’, ‘second tier metro region’, ‘smaller metro region’

and ‘other regions’. Areas in these categories are coded respectively 4 through 1, so that

larger scores indicate bigger city environments. Separately, I also fit a dummy for firms in

London, which takes the value 1 if firm is in London NUTS2 area.

Next, I code firms’ locations into a broader urban-rural typology. To do this I use two

different classifications developed by Eurostat and the UK Office for National Statistics

(ONS). The Eurostat typology has four categories: ‘predominantly urban regions’ (coded 4),

‘intermediate regions, close to a city’ (3), ‘intermediate, remote regions’ (2) and

‘predominantly rural regions, close to a city’ (1). The ONS typology has three broad groups,

‘predominantly urban’ (coded 3), ‘significant rural’ (2) and ‘predominantly rural’ (1).

I then fit interaction terms of firms’ % minority ethnic owners/partners with metro

code, the London dummy and the two urban/rural classifications. If London / big cities /

urban areas amplify outcomes for firms, we should expect coefficients of interaction terms to

be positive. If there is a dampening effect, interactions’ point estimates will be negative.


Results are given in Table 9-11, for product innovation, process innovation and turnover

respectively. In each case column 1 fits the base model, column 2 adds metro code and its

interaction with the main variable; column 3 adds the London dummy, and columns 4-5 add

the Eurostat and ONS urban-rural classifications. For innovation models, results are raw

coefficients.

Table 9 about here

Product innovation results, shown in Table 9, generally show little significant

differences from the base model (column 1). The coefficient of metro areas is negative, as is

the interaction term – but this is small and close to zero. By contrast, diverse London firms

have a greater likelihood of innovating compared to other diverse firms – but again, the effect

is not significant. On both urban-rural classifications, the main coefficient of ETEAM turns

positive (although with large standard errors), and the interaction term is negative (significant

at 5% for the ONS classification).

Table 10 about here

Table 10 gives process innovation results. Perhaps surprisingly, bigger cities are

associated with substantially less process innovation, as are diverse city firms – although the

coefficient is around ten times smaller. Conversely, while firms in London are substantially

and significantly more likely to innovate, diverse London firms are less likely to do so

(although the link is not significant). More broadly, firms in more urban areas are linked to

lower process innovation; but results for diverse urban firms vary across classification

(negative and 5% significant for Eurostat, slightly positive for ONS).

Turnover models are given in Table 11, and as before, differ from the innovation

results. City size and position is strongly linked to turnover, and diverse firms in bigger cities

are linked to higher turnover (significant at 10%). Note that when the interaction term is

fitted, the general (non-city) link from diversity to turnover drops and becomes non-

significant. While there is also a positive London-turnover link, however, the diverse

firm*London term has a small negative (non-significant) coefficient. Both urban-rural

classifications indicate a positive link between diverse urban firms and higher turnover – but


neither is significant, and the ONS coefficient is close to zero. When the Eurostat

classification is fitted, the non-urban diversity-turnover link becomes non-significant.

Table 11 about here

Overall, I find some evidence of amplifying and dampening effects, which are

generally stronger and more visible through a London or ‘city’ lens than a broader urban /

rural one. For diverse firms, London has a (weak) positive link to product innovation, but a

negative link to process innovation and turnover levels. This suggests that agglomeration,

competition and discrimination effects may play out differently for different economic

processes. Bigger cities as a whole may dampen innovation for diverse firms, but have a

significantly positive link to levels of turnover. Urban area-diversity connections are

significantly conditioned by the type of classification used.

8. Robustness checks

I run a series of checks to test for potential specification and endogeneity issues.

First, I add in a number of innovation-related controls and re-run (1) for product and

process innovation, using the 2008 cross-section. As discussed in Section 4, the 2008 NBS

contains information on whether firms are planning to invest in R&D; whether they use

university-industry links for R&D purposes; and whether they use specialist networks to

obtain information. The innovation literature suggests all three will have a positive effect on

innovative activity; networking and U-I activity may also influence, and be influenced by top

team composition. Results are given in Table 12. In each case column 1 fits the pooled

sample, column 2 the 2008 cross-section and column 3 cross section plus additional controls.

Table 12 about here

For product innovation, coefficients of ETEAM change sign, so that the share is

positive and quadratic negative. Significant effects of FTEAM drop away in the cross-

section, with and without additional controls. For process innovation, positive effects of


ETEAM remain in the cross-section, but disappear once additional controls are added. There

is little change for FTEAM, although the sign of the quadratic changes to negative. This

suggests that the innovation results are conditioned by the additional elements included here.

Next, I re-run the turnover models including innovation variables on the right hand

side. Intuitively, successfully commercialised innovative activity should feed through into

greater market share, and thus higher revenue. Table 13 shows the results. Point estimates for

both ETEAM and FTEAM change slightly, but the overall pattern of the main results stays

unchanged.

Table 13 about here

As a further check on the turnover models, I refit the model for 2008 data using more

detailed seven-band turnover information. The rich information on the left hand side of the

model might reduce or amplify the observed diversity effects. The results are given in table

14: column 1 fits the pooled sample, column 2 the 2008 data and column 3 the 2008 data with

seven-band turnover. Fitting the more detailed turnover information does not change the

overall shape of significance levels of the results, although coefficients for individual

variables of interest get bigger.

Table 14 about here

Following this, I re-run all the main models including right-hand side controls for

inputs and sales geographies. As discussed in Section 5, both variables may shape firms’

innovative capacity, business performance and the composition of senior management; not

including them in (1) may omit an important intervening variable. Table 15 presents results

including these controls. The top panel shows results for product innovation. For ETEAM

and FTEAM there is little change.

The middle panel shows selected results for process innovation. Only a couple of the

logit models converge, suggesting that the loss of observations is critical in these cases.

(Results for OLS models, available on request, suggest that fitting both controls together

shrinks the coefficient of ethnic-diverse firms, and renders it insignificant (from 10%

significance); by contrast, the beta of minority-headed firms switches from positive to


negative, and becomes marginally significant. Coefficients of FTEAM are essentially

unchanged.

The bottom panel shows results for turnover. Here, including input and sales controls

amplifies the main results. For ETEAM, fitting both new controls together raises the

coefficient of ethnic diversity firms from 0.165 to 0.218; for FTEAM, the negative

coefficient of female-headed firms switches from -0.280 to -0.325. In both cases results

remain 1% significant.

Table 15 about here

Next I explore whether small and young firms shape the results in ways not captured

by my existing control structure. The literature suggests that such firms play a critical role in

employment growth; and this may well affect turnover as well (Haltiwanger et al., 2010). We

might also expect small, young firms which are diverse or minority/female headed to display

distinctive trajectories – depending on whether positive or negative demographic externalities

predominate. To test this, I build two ‘small and young’ dummies taking the value 1 if firms

are both less than 5 years old and a) are small businesses, with 10-50 employees, or b) are

microbusinesses, with under 10 employees.

Results for innovation models show very little difference to the main findings (full

tables are available on request). The one shift is for process innovation: when the

microbusiness dummy is interacted with ethnic diverse teams, the coefficient of the latter

rises from 0384 to 0.438, significant at 10%. Coefficients of microbusinesses and diverse

microbusinesses are slightly negative, likely reflecting constraints on very small firms.

Table 16 about here

Table 16 presents results for the turnover models. Being a young small business has

no significant link to turnover (column 2); not surprisingly, being a microbusiness attracts a

substantial penalty (column 4). Columns 3 and 5 interact these dummies with the ethnic-

diverse top team dummy. I find a weak negative link to turnover for small, young ethnic

diverse firms (column 3) and for microbusinesses (column 5) but neither is significant.


Finally, I repeat the main results adding in NUTS3-level workforce composition

controls, drawn from the Annual Population Survey for England and Wales. Specifically, I fit

sequentially the a) share of directors, managers and senior officials employed in the NUTS3

working-age population, which functions as a measure of the pool of TMT personnel; b) a

measure of workforce skills, the share of NUTS3 working-age population with degree-level

qualifications; and c) the share of degree holders / senior and management employees. These

are designed both to provide additional area-level information, and to proxy for the firm-level

human capital information not present in the NBS. As such they are less precise than one

would wish, and some coefficients are fitted quite imprecisely. The APS does not cover

Northern Ireland: these firms are dropped from the checks.

Table 17 about here

Results are given in Table 17. The top panel shows results for product innovation;

there is little change when the extra controls are fitted. The middle panel shows results for

process innovation; fitting the area-level skills control slightly raises the beta of ethnic

diverse firms (from 0.384 to 0.415) and makes it significant at 10%. The bottom panel fits

turnover models. Here, coefficients of ethnic diverse teams get larger but also shift from 5 to

10% significance; the beta of minority ethnic-headed firms also gets large and becomes

marginally significant. As with the other two dependent variables, there is very little change

to measures of FTEAM.

9. Discussion

This paper explores the connections between top team ethnic and gender composition,

innovation and revenue levels at the firm level, using a rich dataset of English firms. The

paper makes a number of contributions to the small, but growing literature on dynamic

effects of diversity, co-ethnicity and gender composition on business performance. It is one

of very few firm-level European studies, and is (as far as I am aware) the first of its kind in

the UK.


The early results throw up four headline findings. First, I find evidence that suggests a

non-linear relationship link between ethnic and gender diversity and measures of business

performance. This is in line with theory and some existing empirical evidence.

Second, the sign and strength of the link differs across outcomes. For innovation

models, there is a strong and robust link between ethnic div and process innovation, though

none for product innovation. Joint effect is small net positive, suggesting that positive

affordances of diversity on innovation (ideas pooling, knowledge spillovers) outweigh any

negatives (lower trust and social capital, discrimination). I find no links for gender diversity.

For turnover models, I find strong, significant joint effects for both ethnic and gender

diversity, the former larger than the latter. However, in contrast with the innovation results,

both links are small net negative, suggesting that internal / external demographic constraints

outweigh any positives.

Third, distinguishing between diverse and minority/female-headed businesses is

important to explain these results. For process innovation, I find positive links to ethnically

diverse firms but none to minority-headed firms. For turnover, I find positive to ethnic and

gender-diverse firms, but negative links to minority and female-headed businesses. This

suggests that while business diversity has (internal and external) benefits, being ‘too diverse’

is actually an issue of external constraints to the firm (such as discrimination) rather than

internal problems.

Finally, in line with theory I find some evidence that city and urban form have

amplifying / dampening effects on diversity-business performance links. These intervening

effects are all fairly weak, reflecting the relatively broad-brush approach to spatial

classification. A larger sample with post-code data for firms would allow much more precise

estimates of city and urban effects, and future research could usefully explore this approach.

Extensions and robustness checks suggest two further channels that may be

influencing these results, particularly for turnover models. These are first, the geography of

inputs and sales; and second, specific constraints for small, young businesses. As expected,

including controls on firms' R&D and networking activity also helps explain the main results.


The persistence of the main results suggests robust associations between top team

demographics, process innovation and levels of turnover. However, at this stage these results

cannot be interpreted as causal – because I cannot observe firms’ reactions to any diversity or

sameness ‘effects’. In general, identifying causal effects of firm composition is beset with

challenges, not least because similar firms are likely adopt heterogeneous strategies to deal

with identical management issues (Adams et al., 2010). As Adams et al (ibid) point out,

‘there are no cure-all instruments that one can use to deal with this endogeneity … causality,

in the usual sense, is often impossible to determine.’ (p 97).

Further research could pursue a number of different avenues. First, and most crucially

for UK businesses and policymakers, future studies need to use instruments or other

identification techniques that can identify causal effects of diversity and sameness on firm-

level outcomes as far as possible. Second, differences between top team and wider workforce

demographics-performance channels need better exploration, ideally through large, rich

employer-employee datasets. Third, as noted above, better geo-coded data would allow

clearer identification of city and urban-level intervening factors. Working with large public

datasets and matching across microdata, or pursuing ‘big data’ strategies are both promising

ways forward.

[9148 words]


Tables

Table 1. Summary statistics.

Variable Obs Mean Std.

Dev. Min Max

new product innovation in last 12 months 6235 0.239 0.426 0 1

new process innovation in last 12 months 6235 0.088 0.284 0 1

turnover at site in 4 bands 6235 2.039 0.832 1 4

% minority ethnic owners/partners/directors 6235 0.030 0.159 0 1

(% minority ethnic owners/partners/directors)2 6235 0.026 0.151 0 1

minority ethnic-diverse firm 6235 0.018 0.132 0 1

minority ethnic-headed firm 6235 0.023 0.149 0 1

% female owners/partners/directors 6235 0.259 0.332 0 1

(% female owners/partners/directors)2 6235 0.177 0.296 0 1

minority female-diverse firm 6235 0.346 0.476 0 1

minority female-headed firm 6235 0.099 0.298 0 1

number of owners/partners/directors 6227 2.1 3.6 1 100

no of employees who receive a salary (excl. owners) 6235 25.9 374.9 0 20000

years business in operation 6226 3.478 0.807 1 4

firm is subsidiary of uk parent 6235 0.028 0.166 0 1

firm is subsidiary of foreign parent 6235 0.015 0.123 0 1

firm is ultimate holding company 6235 0.042 0.200 0 1

firm is independent 6235 0.688 0.463 0 1

firm is LLP 6235 0.077 0.267 0 1

business provided some training in past 12 months 6235 0.281 0.450 0 1

growth plan dummy 6051 0.333 0.471 0 1

business is operating below capacity 6235 0.684 0.465 0 1

share of foreign sales banded 5421 3.245 2.726 0 6

share of foreign inputs banded 6235 4.044 2.592 0 6

firm expects to do R&D investment in next 12 months 2894 0.621 0.485 0 1

business uses U-I links for R&D 1734 0.196 0.397 0 1

business uses specialist networks for info 2169 0.416 0.493 0 1

Source: RDA NBS.

Note: ownership truncated to 100 owners / firm.


Table 2. Correlation matrix of main variables.

prodin procin turnover ethownsh eth~h_sq eth_div eth_head femowns

h

fem~h_s

q fem_div

fem_hea

d

new product innovation in last 12 months 1

new process innovation in last 12 months -0.1744 1

turnover at site in 4 bands 0.1251 0.0729 1

% ethnic owners/partners/directors 0.0066 0.0142 -0.0278 1

(% ethnic owners/partners/directors)2 0.0042 0.0097 -0.0364 0.9828 1

minority ethnic-diverse firm 0.0135 0.0227 0.0464 0.318 0.1429 1

minority ethnic-headed firm 0.0002 0.0055 -0.0421 0.9331 0.9822 -0.0205 1

% female owners/partners/directors 0.0042 0.0179 -0.1663 0.0164 0.0152 0.0131 0.0144 1

(% female owners/partners/directors)2 -0.0051 0.0167 -0.1955 0.0133 0.0167 -0.0125 0.0197 0.9398 1

female-diverse firm 0.0294 0.0136 0.0467 0.0118 -0.0012 0.0793 -0.0115 0.4505 0.1224 1

female-headed firm -0.0099 0.0107 -0.187 0.0071 0.015 -0.0402 0.0216 0.7391 0.9185 -0.2405 1

Source: RDA NBS.

Note: Obs = 6235. Correlation matrices for main variables and full controls are available on request.


Table 3. Product innovation. Logistic models.

(1) (2) (3) (4) (5)

% ethnic 0.103 0.060 0.538 -0.366 -0.545

owners/partners/directors (0.189) (0.226) (0.695) (0.779) (0.763)

ethownsh_sq

-0.465 0.455 0.639

(0.661) (0.765) (0.753)

% female

0.623**

owners/partners/directors

(0.256)

femownsh_sq

-0.618*

(0.319)

number of

0.009

0.009 0.008


(0.007)

(0.007) (0.007)

no of employees who receive

0.000

0.000 0.000

a salary (excluding owners)

(0.000)

(0.000) (0.000)

years business in operation

-0.068

-0.067 -0.071

(0.049)

(0.049) (0.048)

firm is subsidiary of uk parent

0.438**

0.440** 0.463**

(0.183)

(0.182) (0.180)

firm is subsidiary of foreign

0.568**

0.567** 0.610**

parent

(0.265)

(0.266) (0.268)

firm is ultimate holding company

0.363**

0.363** 0.370**

(0.159)

(0.159) (0.161)

firm is independent

0.116*

0.116* 0.122*

(0.070)

(0.071) (0.072)

firm is LLP

0.192

0.193 0.194

(0.140)

(0.141) (0.141)

business provided some training

0.248***

0.248*** 0.245***

in past 12 months

(0.057)

(0.057) (0.057)

growth plan dummy

0.809***

0.810*** 0.808***

(0.094)

(0.095) (0.097)

business is operating below

0.111

0.110 0.112

capacity

(0.072)

(0.072) (0.072)

Observations 6203 5885 6203 5885 5885

Log-likelihood -3144.216 -2893.196 -3144.125 -2893.121 -2890.859

Source. RDA NBS. All models use year, sic2 and nuts3 dummies. Raw coefficients. HAC standard errors

clustered on 2-digit sector. Constant not shown. * = result significant at 10%, ** = 5%, *** = 1%.


Table 4. Process innovation. Logistic models.

(1) (2) (3) (4) (5)

% ethnic 0.365* 0.408* 1.901* 1.923* 1.937*


ethownsh_sq

-1.656 -1.637 -1.651

(1.115) (1.178) (1.181)

% female

0.030


(0.308)

femownsh_sq

0.205

(0.338)

number of

0.008

0.007 0.007


(0.005)

(0.005) (0.005)


0.000

0.000 0.000


(0.000)

(0.000) (0.000)


-0.032

-0.034 -0.024

(0.069)

(0.069) (0.072)


0.326

0.319 0.326

(0.257)

(0.258) (0.256)


-0.059

-0.052 -0.022

parent

(0.381)

(0.382) (0.388)


-0.016

-0.017 0.003

(0.197)

(0.196) (0.196)

firm is independent

0.085

0.084 0.088

(0.138)

(0.137) (0.137)

firm is LLP

0.268

0.264 0.277

(0.267)

(0.266) (0.267)


0.387***

0.388*** 0.387***

in past 12 months

(0.108)

(0.109) (0.109)

growth plan dummy

0.825***

0.822*** 0.826***

(0.107)

(0.107) (0.106)


-0.083

-0.082 -0.074

capacity

(0.083)

(0.083) (0.082)

Observations 6139 5831 6139 5831 5831

Log-likelihood -1759.031 -1627.992 -1758.332 -1627.354 -1626.345


clustered on 2-digit sector. Constant not shown. * = result significant at 10%, ** = 5%, *** = 1%.


Table 5. Innovation models. OLS results.

Product innovation Process innovation

(1) (2) (3) (1) (2) (3)

% ethnic owners/partners/directors 0.095 -0.071 -0.098 0.193 0.201 0.203

(0.130) (0.135) (0.131) (0.130) (0.140) (0.140)

ethownsh_sq -0.086 0.084 0.111 -0.173 -0.177 -0.178

(0.123) (0.135) (0.131) (0.134) (0.143) (0.144)

% female owners/partners/directors

0.096**

0.006

(0.040)

(0.024)

femownsh_sq

-0.096*

0.012

(0.051)

(0.028)

Controls N Y Y N Y Y

Observations 6235 5922 5922 6235 5922 5922

r2 0.088 0.122 0.122 0.034 0.057 0.057

Source. RDA NBS. All models use year, sic2 and nuts3 dummies. HAC standard errors clustered on 2-digit sector. Controls as in Tables 3 and 4. Constant not shown.

* = significant at 10%, ** = 5%, *** = 1%.


Table 6. Turnover model. OLS results.

(1) (2) (3) (4) (5)

% ethnic -0.110 -0.011 1.002*** 0.863*** 0.730**


ethownsh_sq

-1.186*** -0.931*** -0.798***

(0.354) (0.293) (0.296)

% female

0.285***


(0.090)

femownsh_sq

-0.602***

(0.083)

number of

0.012***

0.012*** 0.011***


(0.004)

(0.004) (0.004)


0.000***

0.000*** 0.000***


(0.000)

(0.000) (0.000)


0.232***

0.232*** 0.217***

(0.017)

(0.017) (0.018)


0.383***

0.380*** 0.386***

(0.088)

(0.086) (0.087)


0.830***

0.834*** 0.819***

parent

(0.092)

(0.092) (0.092)


0.311***

0.311*** 0.292***

(0.059)

(0.058) (0.059)

firm is independent

0.029

0.029 0.028

(0.029)

(0.029) (0.028)

firm is LLP

0.184***

0.183*** 0.168***

(0.034)

(0.034) (0.036)


0.492***

0.492*** 0.492***

in past 12 months

(0.028)

(0.028) (0.027)

growth plan dummy

0.301***

0.300*** 0.296***

(0.022)

(0.022) (0.022)


-0.051**

-0.051** -0.059***

capacity

(0.018)

(0.018) (0.018)

Observations 6235 5922 6235 5922 5922

r2 0.136 0.365 0.138 0.366 0.378

Source. RDA NBS. All models use year, sic2 and nuts3 dummies. HAC standard errors clustered on 2-digit

sector. Constant not shown. * = result significant at 10%, ** = 5%, *** = 1%.


Table 7. Product and process innovation. Testing diversity and sameness.

Product innovation Process innovation

LOGIT OLS LOGIT OLS

minority ethnic-diverse firm -0.083 -0.016 0.384 0.039

(0.184) (0.032) (0.234) (0.028)

minority ethnic-headed firm 0.076 0.010 0.284 0.025

(0.218) (0.037) (0.260) (0.021)

female-diverse firm 0.165*** 0.025*** 0.078 0.006

(0.059) (0.009) (0.092) (0.007)

female-headed firm 0.030 0.004 0.214 0.015

(0.108) (0.018) (0.204) (0.018)

Controls Y Y Y Y

Observations 5885 5922 5831 5922

Log-likelihood -2890.568 -2981.168 -1626.971 -829.346

r2

0.122

0.057


clustered on 2-digit sector. Controls as in Tables 3 and 4. Constant not shown. * = significant at 10%, ** = 5%,

*** = 1%.

Table 8. Turnover. Testing diversity and sameness.

(1)

minority ethnic-diverse firm 0.165**

(0.076)

minority ethnic-headed firm -0.067

(0.051)

female-diverse firm 0.023

(0.024)

female-headed firm -0.280***

(0.026)

Controls Y

Observations 5922

r2 0.376


sector. Controls as in Tables 3 and 4. Constant not shown. * = significant at 10%, ** = 5%, *** = 1%.


Table 9. Big city, London and urban area checks. Product innovation.

(1) (2) (3) (4) (5)

% minority ethnic owners -0.545 -0.446 -0.575 1.640 0.747

/partners/directors (0.763) (0.787) (0.760) (2.348) (1.031)

ethownsh_sq 0.639 0.648 0.563 0.607 0.732

(0.753) (0.757) (0.747) (0.753) (0.802)

Eurostat metro classification

-0.903

( 1-4, 4 = big metro)

(0.724)

% minority ethnic top team X

-0.038


(0.134)

London firm dummy

1.143

(0.744)


0.353

London firm dummy

(0.314)

Eurostat urban/rural classification

-0.301

(1-4, 4 = urban)

(0.241)


-0.565


(0.564)

ONS urban/rural classification

-0.590

(1-3, 3 = urban)

(0.364)


-0.539**


(0.249)

FTEAM Y Y Y Y Y

Controls Y Y Y Y Y

Observations 5885 5885 5885 5885 5623

Log-likelihood -2890.859 -2890.839 -2890.576 -2890.250 -2792.104



*** = 1%.


Table 10. Big city, London and urban area checks. Process innovation.

(1) (2) (3) (4) (5)

% minority ethnic owners 1.937* 2.682** 2.014* 4.378** 1.443


ethownsh_sq -1.651 -1.536 -1.460 -1.684 -1.191

(1.181) (1.168) (1.224) (1.217) (1.122)


-

13.803***

( 1-4, 4 = big metro)

(0.619)


-0.311


(0.209)

London firm dummy

14.009***

(0.411)


-1.008

London firm dummy

(0.635)


-4.601***

(1-4, 4 = urban)

(0.269)


-0.635**


(0.304)


-7.094***

(1-3, 3 = urban)

(0.336)


0.022


(0.320)

FTEAM Y Y Y Y Y

Controls Y Y Y Y Y

Observations 5831 5831 5831 5831 5566

Log-likeilihood -1626.345 -1625.642 -1625.282 -1625.876 -1550.642



*** = 1%.


Table 11. Big city, London and urban area checks. Turnover.

(1) (2) (3) (4) (5)

% minority ethnic owners 0.730** 0.524 0.744** 0.242 0.768**


ethownsh_sq -0.798*** -0.824*** -0.776** -0.793*** -0.875***

(0.296) (0.294) (0.291) (0.297) (0.295)


0.321***

( 1-4, 4 = big metro)

(0.115)


0.083*


(0.049)

London firm dummy

0.252

(0.206)


-0.146

London firm dummy

(0.183)


-0.101

(1-4, 4 = urban)

(0.162)


0.126


(0.123)


-0.022

(1-3, 3 = urban)

(0.108)


0.015


(0.057)

FTEAM Y Y Y Y Y

Controls Y Y Y Y Y

Observations 5922 5922 5922 5922 5655

R2 0.378 0.378 0.378 0.378 0.378


sector. Controls as in Tables 3 and 4. Constant not shown. * = significant at 10%, ** = 5%, *** = 1%.


Table 12. Innovation models: extra variables.

Product innovation (1) (2) (3)

% minority ethnic -0.545 -0.508 0.370

owners/partners/directors (0.763) (1.659) (2.540)

ethownsh_sq 0.639 -0.170 -1.037

(0.753) (1.636) (2.359)

% female owners/partners/directors 0.623** 0.192 0.003

(0.256) (0.456) (0.655)

femownsh_sq -0.618* -0.438 -0.064

(0.319) (0.517) (0.724)

Standard controls Y Y Y

Further controls N N Y

Observations 5885 2798 1425

Log-likelihood -2890.859 -1139.105 -616.872

Process innovation (1) (2) (3)

% minority ethnic 1.937* 4.439** 3.558

owners/partners/directors (1.106) (2.253) (2.497)

ethownsh_sq -1.651 -3.973* -3.676

(1.181) (2.251) (2.453)

% female owners/partners/directors 0.030 0.461 0.934

(0.308) (0.584) (0.719)

femownsh_sq 0.205 -0.346 -0.769

(0.338) (0.661) (0.764)

Standard controls Y Y Y

Further controls N N Y


Log-likelihood -1626.345 -663.748 -380.412

Source. RDA NBS. All models use year, sic2 and nuts3 dummies, plus controls as in Tables 3 and 4. Raw

coefficients. HAC standard errors clustered on 2-digit sector. Constant not shown. * = significant at 10%, ** =

5%, *** = 1%.


Table 13. Turnover models including innovation.

(1) (2) (3) (4)

minority ethnic-diverse firm 0.165** 0.166** 0.163** 0.163**

(0.076) (0.076) (0.076) (0.076)

minority ethnic-headed firm -0.067 -0.068 -0.069 -0.070

(0.051) (0.052) (0.051) (0.051)

female-diverse firm 0.023 0.022 0.023 0.021

(0.024) (0.024) (0.024) (0.024)

female-headed firm -0.280*** -0.280*** -0.281*** -0.281***

(0.026) (0.026) (0.026) (0.026)

new product innovation in last 12

months 0.069***

0.083***

(0.026)

(0.030)

new process innovation in last 12

months 0.060 0.086

(0.053) (0.058)

Observations 5922 5922 5922 5922

r2 0.376 0.377 0.376 0.378

Source. RDA NBS. All models use year, empl, sic2 and nuts3 dummies. HAC standard errors clustered on 2-

digit sector. Controls as in Tables 3 and 4. Constant not shown. * = significant at 10%, ** = 5%, *** = 1%.


Table 14. Turnover models, seven band turnover.

(1) (2) (3)

minority ethnic-diverse firm 0.165** 0.170* 0.357*

(0.076) (0.094) (0.197)

minority ethnic-headed firm -0.067 -0.028 -0.165

(0.051) (0.087) (0.178)

female-diverse firm 0.023 0.055** 0.096**

(0.024) (0.027) (0.048)

female-headed firm -0.280*** -0.281*** -0.591***

(0.026) (0.036) (0.077)

Controls Y Y Y


r2 0.376 0.406 0.426

Source. RDA NBS. All models use year, empl, sic2 and nuts3 dummies. HAC standard errors clustered on 2-

digit sector. Controls as in Tables 3 and 4. Constant not shown. * = significant at 10%, ** = 5%, *** = 1%.


Table 15. Foreign inputs and sales.

Product innovation (1) (2) (3) (4)

minority ethnic-diverse firm -0.083 -0.357 0.003 -0.362

(0.184) (0.330) (0.182) (0.330)

minority ethnic-headed firm 0.076 -0.380 0.158 -0.387

(0.218) (0.406) (0.255) (0.460)

female-diverse firm 0.165*** 0.157*** 0.162*** 0.154***

(0.059) (0.059) (0.057) (0.057)


(0.108) (0.161) (0.128) (0.165)

Share foreign sales N Y N Y

Share foreign inputs N N Y Y

Observations 5885 2464 5129 2315

Log-likelihood -2890.568 -1156.777 -2620.193 -1084.907

Process innovation (1) (2)

minority ethnic-diverse firm 0.384 0.297

(0.234) (0.257)

minority ethnic-headed firm 0.284 0.023

(0.260) (0.295)

female-diverse firm 0.078 0.104

(0.092) (0.100)

female-headed firm 0.214 0.187

(0.204) (0.231)

Share foreign inputs N Y

Observations 5831 5067

Log-likelihood -1626.971 -1457.035

Turnover (1) (2) (3) (4)

minority ethnic-diverse firm 0.165** 0.238*** 0.183** 0.218**

(0.076) (0.088) (0.073) (0.083)

minority ethnic-headed firm -0.067 -0.032 -0.080 -0.025

(0.051) (0.089) (0.048) (0.097)


(0.024) (0.033) (0.024) (0.034)


(0.026) (0.049) (0.027) (0.052)

Share foreign inputs N Y N Y

Share foreign sales N N Y Y

Observations 5922 2523 5165 2368

r2 0.376 0.441 0.379 0.446

Source. RDA NBS. All models use year, sic2 and nuts3 dummies, plus controls as in Tables 3 and 4. Panels 1

and 2 fit raw coefficients. HAC standard errors clustered on 2-digit sector. Constant not shown. * = significant

at 10%, ** = 5%, *** = 1%.


Table 16. Small and young firms: turnover models.

(1) (2) (3) (4) (5)

minority ethnic-diverse firm 0.165** 0.166** 0.173* 0.163** 0.171**

(0.076) (0.076) (0.088) (0.075) (0.085)

minority ethnic-headed firm -0.067 -0.068 -0.068 -0.064 -0.064

(0.051) (0.051) (0.051) (0.051) (0.051)

sysmall

0.045 0.045

(0.059) (0.060)

eth_sysmall

-0.050

(0.159)

symicro

-0.148*** -0.147***

(0.042) (0.044)

eth_symicro

-0.067

(0.170)

FTEAM Y Y Y Y Y

Observations 5922 5922 5922 5922 5922

r2 0.376 0.376 0.376 0.377 0.377

Source: RDA NBS. All models use year, sic4 and nuts3 dummies, plus controls as in Tables 3 and 4. HAC

standard errors clustered on 4-digit sector. Constant not shown. * = significant at 10%, ** = 5%, *** = 1%.


Table 17. Area-level workforce data. Innovation and turnover models.

Product innovation (1) (2) (3) (4)

minority ethnic-diverse firm -0.083 -0.193 -0.195 -0.191

(0.184) (0.247) (0.255) (0.248)


(0.218) (0.206) (0.205) (0.206)

female-diverse firm 0.165*** 0.182*** 0.182*** 0.182***

(0.059) (0.068) (0.068) (0.068)


(0.108) (0.096) (0.098) (0.097)

% all in employment

0.046

managers directors and senior officials

(0.060)

% working age population

-0.005

with NVQ4 or above

(0.059)

managers as share of NVQ4+ working

2.170

age population

(2.060)

Controls Y Y Y Y

Observations 5885 4999 4999 4999

Log-likelihood -2890.568 -2484.279 -2484.687 -2484.002

Process innovation (1) (2) (3) (4)

minority ethnic-diverse firm 0.384 0.403* 0.415* 0.401

(0.234) (0.245) (0.249) (0.245)


(0.260) (0.277) (0.280) (0.277)


(0.092) (0.100) (0.100) (0.100)


(0.204) (0.193) (0.194) (0.192)

% all in employment

-0.049


(0.060)


-0.081

with NVQ4 or above

(0.072)


-1.521

age population

(2.077)

Controls Y Y Y Y

Observations 5831 4951 4951 4951

Log-likelihood -1626.971 -1353.561 -1353.267 -1353.628

Source: RDA NBS. Raw coefficients. All models use year, sic2 and nuts3 dummies, plus controls as in Tables

3 and 4. HAC standard errors clustered on 2-digit sector. Constant not shown. * = significant at 10%, ** = 5%,

*** = 1%.


Turnover (1) (2) (3) (4)

minority ethnic-diverse firm 0.165** 0.173* 0.172* 0.172*

(0.076) (0.092) (0.092) (0.092)

minority ethnic-headed firm -0.067 -0.100* -0.100* -0.101*

(0.051) (0.052) (0.052) (0.052)


(0.024) (0.025) (0.025) (0.025)


(0.026) (0.029) (0.029) (0.029)

% all in employment

-0.012


(0.011)


0.004

with NVQ4 or above

(0.014)


-0.551

age population

(0.358)

Controls Y Y Y Y

Observations 5922 5035 5035 5035

Log-likelihood 0.376 0.370 0.370 0.370

Source: RDA NBS. All models use year, sic2 and nuts3 dummies, plus controls as in Tables 3 and 4. HAC

standard errors clustered on 2-digit sector. Constant not shown. * = significant at 10%, ** = 5%, *** = 1%.


References

ADAMS, R., B., HERMALIN, B. E. & WEISBACH, M. S. 2010. The Role of Boards of

Directors in Corporate Governance: A Conceptual Framework and Survey. Journal of

Economic Literature, 48, 58-107.

AGRAWAL, A., KAPUR, D. & MCHALE, J. 2008. How do spatial and social proximity

influence knowledge flows? Evidence from patent data. Journal of Urban Economics,

64, 258-269.

AITKEN, B. & HARRISON, A. E. 1999. Do Domestic Firms Benefit from Direct Foreign

Investment? Evidence from Venezuela. The American Economic Review, 89, 605-

618.

ALESINA, A. & FERRARA, E. L. 2005. Ethnic Diversity and Economic Performance.

Journal of Economic Literature, 43, 762-800.

ANGRIST, J. & PISCHKE, J.-S. 2009. Mostly Harmless Econometrics, Princeton, Princeton

University Press.

ASHRAF, Q. & GALOR, O. 2011. The "Out of Africa" Hypothesis, Human Genetic

Diversity, and Comparative Economic Development. National Bureau of Economic

Research Working Paper Series, No. 17216.

ASIEDU, E., FREEMAN, J. A. & NTI-ADDAE, A. 2012. Access to Credit by Small

Businesses: How Relevant Are Race, Ethnicity, and Gender? The American Economic

Review, 102, 532-537.

ASPINALL, P. 2009. The Future of Ethnicity Classifications. Journal of Ethnic and

Migration Studies, 35, 1417-1435.

BANTEL, K. A. & JACKSON, S. E. 1989. Top management and innovations in banking:

Does the composition of the top team make a difference? Strategic Management

Journal, 10, 107-124.

BERLIANT, M. & FUJITA, M. 2007. Knowledge creation as a square dance on the Hilbert

cube. MPRA Paper 2884.

BERLIANT, M. & FUJITA, M. 2009. The Dynamics of Knowledge Diversity and Economic

Growth. 56th Annual North American Meeting, Regional Science Association

International. San Francisco.

BIOSCA, A. B., MARSTON, L. & WESTLAKE, S. 2011. Vital Growth: Vital GrowthThe

importance of high-growth businesses to the recovery. London: NESTA.

BOSCHMA, R. 2005. Proximity and Innovation: A Critical Assessment. Regional Studies,

39, 61 - 74.

BUYL, T., BOONE, C., HENDRIKS, W. & MATTHYSSENS, P. 2010. Top Management

Team Functional Diversity and Firm Performance: The Moderating Role of CEO

Characteristics. Journal of Management Studies, 48, 151-177.


CARPENTER, M. A., GELETKANYCZ, M. A. & SANDERS, W. G. 2004. Upper Echelons

Research Revisited: Antecedents, Elements, and Consequences of Top Management

Team Composition. Journal of Management, 30, 749-778.

CASEY, T. & DUSTMANN, C. 2009. Immigrants' Identity, Economic Outcomes and the

Transmission of Identity Across Generations. CReAM Discussion Paper Series CDP

23/09. London: University College London.

CERTO, S. T., LESTER, R. H., DALTON, C. M. & DALTON, D. R. 2006. Top

Management Teams, Strategy and Financial Performance: A Meta-Analytic

Examination. Journal of Management Studies, 43, 813-839.

D’ESTE, P., GUY, F. & IAMMARINO, S. 2011. Shaping the formation of university-

industry research collaborations: what type of proximity does really matter? Papers in

Evolutionary Economic Geography 11.06. Utrecht: Urban and Regional Research

Centre.

DAHLIN, K. B., WEINGART, L. R. & HINDS, P. J. 2005. TEAM DIVERSITY AND

INFORMATION USE. Academy of Management Journal, 48, 1107-1123.

DAHLMAN, C. 2010. Innovation Strategies of Three of the BRICS: Brazil, India and China-

What Can We Learn From Three Different Approaches. In: FU, X. & SOETE, L.

(eds.) The Rise of Technological Power in the South. Basingtoke: Palgrave

MacMillan.

DAVIES REVIEW 2011. Women on Boards. London: Department of Business, Innovation

and Skills.

DEPARTMENT OF INNOVATION UNIVERSITIES AND SKILLS 2008. Innovation

Nation. London: Department of Innovation, Universities and Skills.

DEZSÖ, C. L. & ROSS, D. G. 2012. Does female representation in top management improve

firm performance? A panel data investigation. Strategic Management Journal, 33,

1072-1089.

DOCQUIER, F. & RAPOPORT, H. 2012. Globalization, Brain Drain, and Development.

Journal of Economic Literature, 50, 681-730.

DURANTON, G. & PUGA, D. 2001. Nursery Cities: Urban Diversity, Process Innovation

and the Life Cycle of Products. American Economic Review 91, 1454-1477.

FAGERBERG, J. 2005. Innovation: A guide to the literature. In: FAGERBERG, J.,

MOWERY, D. & NELSON, R. (eds.) The Oxford Handbook of Innovation. Oxford:

OUP.

FOLEY, C. F. & KERR, W. R. 2011. Ethnic Innovation and U.S. Multinational Firm

Activity. National Bureau of Economic Research Working Paper Series. Cambridge,

Mass.: NBER.

FRANCOEUR, C., LABELLE, R. A. & SINCLAIR-DESGAGNÃ©, B. 2008. Gender

Diversity in Corporate Governance and Top Management. Journal of Business Ethics,

81, 83-95.


FUJITA, M. & WEBER, S. 2003. Strategic Immigration Policies and Welfare in

Heterogenous Countries. Institute of Economic Research Working Papers. Kyoto:

Kyoto University.

GOLDIN, I., CAMERON, G. & BALARAJAN, M. 2011. Exceptional People: How

migration shaped our world and will define our future, Oxford, Princeton University

Press.

GORDON, I., WHITEHEAD, C. & TRAVERS, T. 2007. The Impact of Recent Immigration

on the London Economy. London: City of London Corporation.

GREEN , E. 2011. Endogenous Ethnicity. monograph. London: DESTIN, London School of

Economics.

GRIFFITH, R., HUERGO, E., MAIRESSE, J. & PETERS, B. 2006. Innovation and

Productivity Across Four European Countries. Oxford Review of Economic Policy,

22, 483-498.

HALTIWANGER, J. C., JARMIN, R. S. & MIRANDA, J. 2010. Who Creates Jobs? Small

vs. Large vs. Young. NBER Working Paper 16300. Cambridge, MA: NBER.

HAMBRICK, D. C. & MASON, P. A. 1984. Upper Echelons: The Organization as a

Reflection of Its Top Managers. The Academy of Management Review, 9, 193-206.

HARRISON, A. & RODRÍGUEZ-CLARE, A. 2009. Trade, Foreign Investment, and

Industrial Policy for Developing Countries. National Bureau of Economic Research

Working Paper Series No. 15261. Cambridge, MA: NBER.

HART, D. 2010. What Do Foreign-Born Founders Bring To Entrepreneurial Teams? An

Exploration in the U.S. High-Tech Sector. School of Public Policy Working Papers

Arlington, VA: George Mason University.

IPSOS MORI 2009. The National Business Survey: National Report. London: Ipsos MORI.

JACKSON, S. E., JOSHI, A. & ERHARDT, N. L. 2003. Recent Research on Team and

Organizational Diversity: SWOT Analysis and Implications. Journal of Management,

29, 801-830.

JACOBS, J. 1969. The Economy of Cities, London, Vintage.

JAVORCIK, B. S. 2004. Does Foreign Direct Investment Increase the Productivity of

Domestic Firms? In Search of Spillovers Through Backward Linkages. The American

Economic Review, 94, 605-627.

JAVORCIK, B. S., ÖZDEN, Ç., SPATAREANU, M. & NEAGU, C. 2011. Migrant networks

and foreign direct investment. Journal of Development Economics, 94, 231-241.

KAPUR, D. & MCHALE, J. 2005. Sojourns and Software: Internationally mobile human

capital and high tech industry development in India, Ireland and Israel. In: ARORA,

A. & GAMBARDELLA, A. (eds.) From Underdogs to Tigers: The Rise and Growth

of the Software Industry in Brazil, China, India, Ireland and Israel. Oxford: OUP.


KERR, S. P. & KERR, W. 2011. Economic Impacts of Immigration: A Survey NBER

Working Paper 16736. Cambridge, MA: NBER

KERR, W. 2008. Ethnic Scientific Communities and International Technology Diffusion.

Review of Economics and Statistics, 90, 518-537.

LAUREN, K., MAHNKE, V. & VEJRUP-HANSEN, P. 2004. Do differences make a

difference? The impact of human capital diversity, experience and compensation on

firm performance in engineering consulting. DRUID 4th Summer Conference.

Elsinore, Denmark.

LEE, N. 2012. Free to Grow? Assessing the obstacles faced by actual and potential high

growth firms. London: Big Innovation Centre / The Work Foundation.

MARÉ, D. C. & FABLING, R. 2011. Productivity and Local Workforce Composition. In:

11-10, M. W. P. (ed.). Wellington, NZ: Motu Economic and Public Policy Research.

MARÉ, D. C., FABLING, R. & STILLMAN, S. 2011. Immigration and Innovation. IZA

Discussion Paper 5686. Bonn: IZA.

NARANJO-GIL, D., HARTMANN, F. & MAAS, V. S. 2008. Top Management Team

Heterogeneity, Strategic Change and Operational Performance*. British Journal of

Management, 19, 222-234.

NATHAN, M. 2012. After Florida: Towards an Economics of Diversity. European Urban

and Regional Studies, Online First.

NATHAN, M. & LEE, N. Forthcoming. Cultural diversity, innovation and entrepreneurship:

Firm-level evidence from London. Economic Geography.

OFFICE OF NATIONAL STATISTICS 2012a. Ethnicity and National Identity in England

and Wales 2011. Newport: ONS.

OFFICE OF NATIONAL STATISTICS 2012b. International Migrants in England and Wales

2011. Newport: ONS.

OSTERGAARD, C. R., TIMMERMANS, B. & KRISTINSSON, K. 2011. Does a different

view create something new? The effect of employee diversity on innovation.

Research Policy, 40, 500-509.

OTTAVIANO, G., BELLINI, E. & MAGLIETTA, A. 2007. Diversity and the Creative

Capacity of Cities and Regions. SUSDIV Paper 2.2007. Bologna: FEEM.

OZGEN, C., NIJKAMP, P. & POOT, J. 2011. The Impact of Cultural Diversity on

Innovation: Evidence from Dutch Firm-Level Data. IZA Discussion Paper 6000.

Bonn: IZA.

PAGE, S. 2007. The Difference: How the Power of Diversity Creates Better Groups, Firms,

Schools and Societies, Princeton, Princeton University Press.


PARROTTA, P., POZZOLI, D. & PYTLIKOVA, M. 2011. The Nexus Between Labor

Diversity and Firm's Innovation [sic]. NORFACE Discussion Paper 2011-5. The

Hague.

PATACCHINI, E. & ZENOU, Y. 2012. Ethnic Networks and Employment Outcomes.

CREAM Discussion Paper 02/12. London: UCL

PITCHER, P. & SMITH, A. D. 2001. Top Management Team Heterogeneity: Personality,

Power, and Proxies. Organization Science, 12, 1-18.

RODRIK, D. 2004. Industrial Policy for the Twenty-First Century. CEPR Discussion Paper

4767. London: CEPR.

ROMER, P. 1990. Endogenous Technological Change. Journal of Political Economy, 98, 71-

102.

SAXENIAN, A.-L. 2006. The New Argonauts: Regional Advantage in a Global Economy,

Cambridge, MA, Harvard University Press.

SAXENIAN, A.-L. & SABEL, C. 2008. Venture Capital in the ‘Periphery’: The New

Argonauts, Global Search and Local Institution-Building. Economic Geography, 84,

379-394.

SMITH, K. 2005. Measuring innovation. In: FAGERBERG, J., MOWERY, D. & NELSON,

R. (eds.) The Oxford Handbook of Innovation. Oxford: Oxford University Press.

VON HIPPEL, E. 2005. Democratising Innovation, Cambridge, MA, MIT Press.

WIERSEMA, M. F. & BANTEL, K. A. 1992. TOP MANAGEMENT TEAM

DEMOGRAPHY AND CORPORATE STRATEGIC CHANGE. Academy of

Management Journal, 35, 91-121.

WILLIAMS, K. Y. & O'REILLY, C. 1998. Demography and diversity in organisations: a

review of 40 years of research. Research in Organisation Behaviour, 20, 77-140.

YEUNG, H. 2009. Regional Development and the Competitive Dynamics of Global

Production Networks: An East Asian Perspective. Regional Studies, 43, 325-351.

ZENOU, Y. 2011. Spatial versus Social Mismatch: The Strength of Weak Ties. IZA

Discussion Papers 5507. Bonn: Institute for the Study of Labor (IZA).

(Casey and Dustmann, 2009)

(2005).

TOP TEAM DEMOGRAPHICS, INNOVATION AND BUSINESS PERFORMANCE ... · empirical evidence exists...

Documents

Transcript of TOP TEAM DEMOGRAPHICS, INNOVATION AND BUSINESS PERFORMANCE ... · empirical evidence exists...